15

A simple question:

What qualifies as a research publication?

Two criteria spring to mind:

  1. Peer-reviewed: The article has gone through some form of formal peer-review process.
  2. Dead-tree proceedings: The article has been published in a bound collection of thin sheets of dead tree.

In terms of exceptions to 1., many book chapters do not rely on a formal peer-review process but are generally considered as research publications and listed in author bibliographies. Similarly, some may consider arXiv, etc., as publications and counting towards h-index counts.

In terms of exceptions to 2., various workshops (esp. in CS) publish on-line through systems like CEUR or informal web-site proceedings, are peer-reviewed and contain in-depth technical material. As online publishing becomes more commonplace, criterion 2 will grow weaker and weaker.

Is there something else I'm missing?

  • 2
    A poem published in a ("peer-reviewed") poetry magazine satisfies your two criteria, but I don't think it would be considered a research publication. – Joel Reyes Noche Jul 16 '13 at 2:56
  • 2
    @badroit this is a great question and I amazed it hasn't been asked before. would you edit the question and remove the justification part? – seteropere Jul 16 '13 at 7:28
  • While this question is interesting, I'm not sure where it's relevant. Any work can be cited, regardless of whether it's a research publication or not; it's up to the reader to decide whether the source is convincing. The only application I can think of is how to list publications on a resume, and that's typically done in sections; peer-reviewed, proceedings, book chapters, etc. The question is interesting, but largely academic. – eykanal Jul 16 '13 at 14:08
  • 2
    @eykanal, I had a list of criteria why I felt the question was concrete and important (not an exercise metaphysics): (1) related works: what is in scope for surveys, for declarations of novelty, for comparative studies, (2) publication listings: CVs, homepages, publication indexes, (3) author metrics: publication counts, citation counts, h-indexes, (4) self-plagiarism: what is "already published", what can/not be copied from self verbatim, etc. I removed these per request as people felt they had different answers. I rather feel the question generalises all of these issues. – badroit Jul 16 '13 at 14:57
12

Your criteria are close, but (at least for computer science) not quite right.

  1. Research: the paper must describe a novel contribution to some field of inquiry. Whether a paper makes a novel (and correct and interesting) contribution is often determined through formal peer review, but not always. Peer review is almost always a formal requirement for a paper to "count" for purposes of formal evaluation for promotion and tenure, but not necessarily for building one's reputation as a researcher. (It's quite common for breakthrough results in theoretical compute science to be announced and discussed in blogs months before they undergo formal peer review.)

  2. Publication: The paper must be, at least in principle, accessible to the public. Two reasonable prerequisites are that the paper has a DOI, and that it is published either in a serial with an ISSN or in a book with an ISBN, but one could make a reasonable case that PDfs on researchers' personal web pages, blog and newsgroup posts, and even StackExchange answers are "publications". (Certainly there are fields, like chemistry, in which journals will not publish anything that has previously appeared on the web, on the grounds that it is "already published".) There is absolutely no requirement that the paper be printed on dead trees to be considered "real".

In particular, I would classify most ArXiv preprints as "research publications", despite the lack of peer review and the absence of dead trees.

  • Thanks for the comments. Staying within Computer Science, I find that for academic evaluation, even aside from books, invited book chapters with tertiary material are often ranked more highly than hard-fought "A+" conference papers. Similarly, survey journal papers without novel results are often highly cited and highly prized (and no doubt, highly useful aside from having a well-selected title). Hence it seems to me that "novel contributions" (in a technical sense) are not a necessary condition for highly-valued research publications. – badroit Jul 16 '13 at 2:11
  • 3
    ...invited book chapters with tertiary material are often ranked more highly than hard-fought "A+" conference papers. — That is not my experience at all, but every department is different. Also: Well-written synthesis of existing results is a novel contribution. – JeffE Jul 16 '13 at 15:56
  • 1
    can I copy parts of my own arXiv paper verbatim — In my opinion, only if you're writing another version of the same paper. (Recall that different versions of a single paper may be published as a technical report, as an ArXiv preprint, in a conference proceedings, in a journal, and as a thesis chapter.) – JeffE Jul 16 '13 at 16:00
  • 1
    need to discuss and compare to claims that are not peer-reviewed — Somebody else needs to handle that one. I've never understood how systems performance claims are (or can be, even in principle) peer-reviewed. – JeffE Jul 16 '13 at 16:01
  • 1
    I've never understood how systems performance claims are (or can be, even in principle) peer-reviewed ... Probably a good candidate for a separate question, but the key aspects are reproducibility of methods described and consistency of results. Although common and even encouraged, I dislike first-party system comparisons since there is an obvious bias of authors towards their own system. Third-party evaluations or standardised benchmarking schemes work much better and there are often separate tracks for this at conferences. – badroit Jul 18 '13 at 17:27
4

The Committee on Publication Ethics (COPE) has summarized a standard for responsible publishing for authors (pdf). Although this document does not define scientific publication explicitly, the requirements are implicitly understood by reading about the demands set on an author of such a publication.

The Council of Science Editors (CSE) (quoted by Day & Gastel, 2012, p 19) provides the following definition

An acceptable primary scientific publication must be the first disclosure containing sufficient information to enable peers (1) to assess observations, (2) to repeat experiments, and (3) to evaluate intellectual processes; moreover it must be susceptible to sensory perception, essentially permanent, available to the scientific community without restriction, and available for regular screening by one or more of the major recognized secondary service.

This means that, for example, abstracts, corporate reports do not count as scientific publications. It should be added that review papers of course count although the wording of the definition may not make that clear at first.

Day, R.A. & Gastel, B., 2012 How to write and publish a scientific paper. Seventh Edition. Cambridge, Cambridge University Press.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.